融合小波包和神经网络的脑电信号处理方法

针对运动想象脑电信号处理中分类准确率较低的问题，提出了一种基于能量（二阶矩）小波包变换和莱文伯格-马夸特神经网络算法相结合的运动想象脑电信号处理方法.首先，利用能量方法对信号进行时域分析，选取有效的时序段；然后，使用小波包变换对所选有效时域段的各导信号进行时频分解，选取与想象任务相关的频段信息重构脑电信号特征；最后，将各导信号重构的特征串接，导入基于莱文伯格-马夸特训练算法的神经网络实现最终的任务分类.利用2个脑电信号标准竞赛数据库进行方法验证，分别取得了95.62%和90.13%的分类准确率.与近期的一些研究成果进行对比，可知该方法具有较好的分类效果.

An Electroencephalogram Signal Processing Method Fusing Wavelet Packet and Neural Network

Aiming at the classification accuracy of the motor imagery electroencephalogram in processing is low, a processing method based on the combination of energy (second-order moment) wavelet packet transform and Levenberg-Marquardt neural network is proposed. Firstly, the energy method is used to analyze signal in the time domain, and the effective time sequence is selected. Then, wavelet packet transform is used to decompose the time-frequency of each pilot signal in the selected effective time-domain segment, and the frequency information related to the imagination task is selected to reconstruct the signal characteristics. Finally, the features reconstructed by each guide signal are concatenated and imported into the neural network based on the Levenberg-Marquardt training algorithm to realize the task classification. The method was verified by two kinds of electroencephalogram signal standard competition database, and the classification accuracy is 95.62% and 90.13%, respectively. Compared with some recent research results, this algorithm has a better processing effect.